OcuNet - Eye Disease Classification Model

Model Description

OcuNet is a deep learning model for classifying retinal fundus images into four categories: Cataract, Diabetic Retinopathy, Glaucoma, and Normal.

• Architecture: EfficientNet-B0 + Custom Classification Head
• Parameters: 4,665,472
• Input: 224×224 RGB images
• Output: 4-class probability distribution

Intended Use

• Primary screening for eye diseases
• Clinical decision support
• Telemedicine applications
• Educational purposes

Performance

Metric	Score
Accuracy	86.89%
ROC-AUC	97.88%
F1 (Macro)	86.79%

Limitations

• Trained on limited dataset (4,217 images)
• Lower accuracy on "Normal" class (72.67% recall)
• No severity grading
• Requires clinical verification

Training Data

Eye Diseases Classification dataset from Kaggle with 4,217 fundus images across 4 classes.

How to Use

from predict import EyeDiseaseClassifier

classifier = EyeDiseaseClassifier()
result = classifier.predict("image.jpg")
print(result['predicted_class'], result['confidence'])